Landslides (2016) 13:1181–1196DOI 10.1007/s10346-015-0640-zReceived: 23 April 2015Accepted: 14 September 2015Published online: 10 October 2015© Springer-Verlag Berlin Heidelberg 2015

A. Schimmel I J. Hübl

Automatic detection of debris flows and debris floodsbased on a combination of infrasound and seismicsignals

Abstract Processes like landslides, debris flows, or bed loadtransport, at the intersection between the natural environmentand human activity, constitute an increasing threat to peopleand property. The ability to detect these processes prematurelyis an essential task for mitigating these hazards. Past studieshave shown that debris flows and debris floods emit detectablesignals in the low-frequency infrasonic spectrum and inducecharacteristically seismic signals. A number of monitoring de-vices and detection methods to identify debris flows using thesesignals have been developed, but up to date, no warning systembased on a combination of seismic and infrasound sensors hasbeen considered. Previous studies have already shown that seis-mic and infrasonic signals of alpine mass movements are corre-lated and complementary and that the combination of these twosensor types can serve as basis for an error-resistant detectionand warning system. So this work aims to develop a detectionsystem for detecting debris flows and debris floods by analyzingthe seismic and infrasound waves. The system is build up on aminimum of one seismic and one infrasound sensor which areco-located and a microcontroller which runs a detection algo-rithm to detect debris flows and debris floods with high accu-racy in real time directly on-site. The detection algorithm isbased on an analysis of the evolution in time of the frequencycontent of the mass movement signal and has been tested withdebris flows and debris flood signals monitored at different testsites in Austria and Switzerland. This paper describes the cur-rent version of the detection system and gives an example ofevent detection at the Tyrolese test sites Lattenbach, Dristenau,and Farstrinne.

Keywords Debris flows . Debris floods . Early detectionsystem . Infrasound signals . Seismic signals

IntroductionEarly warning systems for alpine hazards are essential in order todecrease the potential threat to humans and buildings. Monitoringsystems based on seismic or infrasound signals are quite commonand have been used to study alpine mass movements like debrisflows, debris floods, and avalanches for many years. For monitor-ing purposes, both seismic waves as well as infrasonic waves havebenefits and drawbacks. Seismic waves can be divided into bodywaves (frequency 0.02–100 Hz, velocity up to 5,000 m/s in granite)and surface waves (frequency 0.003–0.1 Hz) which travel moreslowly when they are located directly under the surface. Bodywaves, travelling through the interior of the earth, can further bedivided into primary waves (longitudinal or pressure waves) andsecondary waves (transverse waves). They vary according to den-sity and modulus (stiffness) of the ground. The distance betweenthe sensor and the mass movement and the characteristics of thesite has a strong influence on the registered seismic signals

(Biescas et al. 2003). The main source of the seismic energygenerated by debris flow is the basal friction of the dense bodyinside the flow in contact with the ground (e.g., Arattano 2003;Biescas et al. 2003).

Infrasound signals (0.01–20 Hz) are longitudinal pressurewaves travelling through the air with a velocity of 343 m/s (stan-dard temperature and pressure (STP)), which is approximatelythat of audible sound. Mass movement-generated infrasound sig-nals have a specific amplitude and occupy a relatively noise-freeband in the low-frequency acoustic spectrum. Infrasound cantravel thousands of kilometers and remain detectable over suchdistances. This is due to the frequency dependency of atmosphericattenuation, absorbing high-frequency (audible and ultra-) soundmore than low-frequency (infra-) sound (Pilger and Bittner 2009).Infrasound signals of debris flows are expected to be produced bythe violent surge front and the collision (or abrasion) betweendebris flow and the channel loose boundary (Chou et al. 2007,2010; Kogelnig et al. 2010).

Various previous studies on debris flows (e.g., Wu et al. 1990;Marchi et al. 2002; Arattano 2003; Huang et al. 2003, 2007;Vilajosana et al. 2008; Coviello et al. 2015) and avalanches (e.g.,Suriñach et al. 2000; Biescas et al. 2003) have already shown that itis possible to detect and monitor these processes with geophonesand that it is possible to distinguish them from other seismicsources. The first attempts of infrasound monitoring of debrisflow (Chou et al. 2007; 2010; Kogelnig et al. 2008; Kogelnig et al.2009; Zhang et al. 2004), landslides (Bedard 1996), and avalanches(Bedard 1989, 1994; Scott 2004, 2006; Scott et al. 2007; Sommerfeld1977; Sommerfeld and Gubler 1983) have already proven the via-bility of infrasonic waves in the detection and monitoring of thesetypes of mass movements. However, the potential combination ofinfrasonic and seismic sensors for monitoring natural hazards hasonly been evaluated rarely (Hübl et al. 2041; Kogelnig et al. 2010,2012; Suriñach et al. 2009), and no automatic event detectionbased on a combination of those signals has been developed todate. This combination can take advantages of both sensor tech-nologies and minimized disadvantages (e.g., seismic: lower distur-bances due to wind and weather but strong dependency on thegeology of the site and high attenuation with increasing distancebetween mass movement and sensor; infrasound: little attenuationin the air at local distances but high background noise induced bywind).

This paper presents a new approach to early detection sys-tems and hazard monitoring based on this combination ofseismic and infrasound sensors. The benefits of this systeminclude independence from weather conditions with regard tovisibility, no structural need for sustainability, same system forsnow avalanches (Schimmel et al. 2013a) and debris flows/debrisfloods, and monitoring from a remote location unaffected bythe process.

Landslides 13 & (2016) 1181

Original Paper

Detection system

System setupThe main idea for the setup is to build up a very simple systembased on the infrasound and seismic sensor (Fig. 1) with dataprocessing in real time directly at the sensor site, which can beeasily installed near a torrent and can offer a cost-efficient solutionfor early warning. The system presented in this paper acts as adetection system for debris flow and debris flood which can beenhanced to an early warning system. Therefore, the system has tobe tested in a long-term use regarding its stability and extended tocommunicate the threat (e.g., radio control, SMS alarm, etc.). Thearea of application of such a warning system could be, e.g., theprotection of roads and railways by controlling a traffic light or toget information of the frequency of alpine mass movements toassist regional or local authorities whose task is to actively reducethe risk of such hazards. The easy installation of such a system willalso offer a good solution for protecting workers at constructionsites inside torrents as who are clearing the basin after a debrisflow event.

Currently, two different types of infrasound sensors and twodifferent geophones are used for this system. The mainly usedinfrasound sensor is the Chaparral Physics Model 24 which has aresolution of 2 V/Pa and a frequency range of 0.1 to 100 Hz. Theseismic signal is measured by a geophone of the type Sercel SG-5with a resolution of 80 V/m/s and a natural frequency of 5 Hz.Additionally, the infrasound microphone MK-224 with a resolu-tion of 50 mV/Pa and a frequency range of 3 to 200 Hz as well asgeophones of the type SENSOR SM-4 with resolution of 28.8 V/m/sand natural frequency of 10 Hz are used. The data processing isdone by a Stellaris Luminary Evaluation board LM3S8962 with a50-MHz ARM Cortex-M3 microprocessor. The input signal isadapted to the microcontroller ADC input range by an operationalamplifier circuit. For this input circuit, an inverting amplifier witha gain of 0.2 for the infrasound sensor Chaparral Physics Model 24and a gain of 8 for the MK-224 is used, which results in a finalresolution of 400 mV/Pa for the infrasound signal. The seismicsignal is converted with a gain of 100 which led to a resolution of8 mV/μm/s for the SG-5 sensors and an overall resolution of2.88 mV/μm/s for the SM-4 sensors. The input from the infrasound

sensor is filtered by a high-pass filter based on a RC circuit with acutoff frequency below 1 Hz to eliminate the constant component.The signals are recorded by the microcontroller at a sample rate of100 samples/s with a 32× hardware oversampling to avoid aliasingeffects. The evaluation board offers the possibility to store the dataon a micro-SD card with a size up to 16 GB, and it also providesethernet access for remote control and download of data. The datais written in hour-files and a 16-GB card offers space for 3,560 files,which means a storage time for up to 4 months. A great advantageof this system is that the microcontroller makes it very flexible andadaptable for the special requirements of a detection system andthe setup is low cost, highly effective, and applicable in harshalpine environments. Also, the energy consumption of this systemis very low which makes this setup very useful for standalonestations with solar power supply. The setup works with a powersupply of 12 V DC and needs an electrical power below 1.5 W.

In the summer season of 2013, five test sites in Austria wereequipped with this system, whereby four stations used theChaparral Physics Model 24 and one station (Dristenau) wasequipped with the MK-224 infrasound sensor. Since the SG-5geophones were purchased in August 2013, the SM-4 geophonewas first used for all stations, and two of the stations(Dristenau, Farstrinne) were equipped with SG-5 geophonesfor the second half of the season. In summer 2014, theinfrasound setup has not been modified, but the SG-5 geo-phones were used for all stations. The Wartschenbach test sitewas not in operation in this season, and the system atSchüsserbach worked due to power supply problems only forthe first half of the summer. A view of the test sites locations isgiven at the map in Fig. 2, and Table 1 shows the times ofoperation and an overview of the used equipment.

For a proper functionality of the detection system, the installa-tion sites have to fulfill some requirements: An ideal system setupis close to the torrent protected from wind (e.g., dense forest) andother sources of interference like cars, trains, etc., and the site hasa consolidated soil or rock for the installation of the seismicsensor. If there is no other possibility for power supply, a solarpower supply is used; the site has to get sufficient hours of solarradiation per day. In an optimal way, there should also be thepossibility for remote control via a mobile network.

Fig. 1 System overview

Original Paper

Landslides 13 & (2016)1182

Detection algorithmThis section describes the current version of a detection method toautomatically detect debris flows and debris floods based onseismic and infrasound data. The requirement on this detectionalgorithm is to identify events as early as possible without manyfalse alarms in an uncomplex way so that the algorithm can be runin real time directly at the sensor site without high computationaleffort (e.g., on a microcontroller). This leads to an approach ofanalyzing the development of the amplitudes of the signals in atime-frequency range. The automatic detection of an event islimited by a minimal event size, weather condition, distance, andbackground noise. The system has two different levels for thedetection output which are based on the amplitudes of theinfrasound and seismic signal, to distinguish between small andlarger events. The infrasound signal and the seismic signal areprocessed by fast Fourier transform (FFT) every second and ana-lyzed with respect to time, time-frequency, and amplitude. For theFFT, the Bluestein FFT algorithm (also called the chirp z-transformalgorithm; Rabiner et al. 1969) is used with 100 samples. Thisalgorithm computes the discrete Fourier transform (DFT) ofnon-power-of-two sizes by re-expressing the DFT as a convolution.

Different methods are used for the event identification basedon seismic and infrasonic signals. For the infrasound signals, thedetection algorithm compares the development of the signal overtime in four frequency bands. These frequency bands were chosento represent the whole characteristic spectrum of infrasound sig-nals produced by debris flows and debris floods (e.g., Kogelnig2012, Hübl et al. 2013) whereby a lower frequency band is used fordebris flow and a higher one for debris flood. The infrasoundsignals of debris flows/debris floods present a typical divergenceover time within these frequency bands, so the average amplitudesof these different frequency bands can be used as detectioncriteria. In the current version of the detection criteria, the averageamplitudes of the debris flow/debris flood frequency bands(avAmpDFlow, avAmpDFlood) have to be beyond the average ampli-tudes of the band below and above (avAmplow, avAmphigh), andthey have to exceed a certain limit. To distinguish between differ-ent event sizes, two limits are used for the average amplitudes, alimit of 10 Pa to detect even small events (level 1 (L1): AmpLimitL1)and a limit of 30 Pa for events with higher magnitude (level 2 (L2):AmpLimitL2). These limits can be changed depending on theapplication of the detection system and the background noise.

Table 1 Overview of the equipment at the test sites and operation times

Test site Infrasound sensor Seismic sensor Start End Operating hours

Lattenbach Chaparral M24 SM-4 5 June 2013 27 December 2013 4,915

Lattenbach Chaparral M24 SG-5 30 April 2014 14 January 2015 6,184

Farstrinne Chaparral M24 SM-4/SG-5 09 July 2013 21 October 2013 2,490

Farstrinne Chaparral M24 SG-5 30 April 2014 15 October 2014 4,026

Wartschenbach Chaparral M24 SM-4 04 July 2013 16 September 2013 1,771

Dristenau MK-224 SM-4/SG-5 06 June 2013 28 September 2013 2,740

Dristenau MK-224 SG-5 16 May 2014 24 September 2014 3,141

Schüsserbach Chaparral M24 SM-4 01 August 2013 13 September 2013 1,025

Schüsserbach Chaparral M24 SG-5 21 May 2014 17 July 2014 1,364

Fig. 2 Map of the test sites equipped with the warning system since 2013

Landslides 13 & (2016) 1183

As an additional criterion, the variance of the amplitudes of theseismic and infrasound signals can be used. Since the variance ofan artificial noise is high compared to the variance of the broad-band debris flow or debris flood signal because in most cases theyare produced by narrow-band sources, these criteria can be usedto eliminate artificial interfering signals. For a detection, the var-iance of the amplitudes of the debris flow/flood frequency band(AmpVarDFlow, AmpVarDFlood) has to be under a certain limit(VarLimit) to ensure that this is a natural signal.

These detection criteria for the infrasound signals are shown bythe following formulae:

Amplitude criteria—level 1:

avAmpDFlow ≥AmpLimitL1 or avAmpDFlood≥AmpLimitL1 ð1Þ

Amplitude criteria—level 2:

avAmpDFlow ≥AmpLimitL2 or avAmpDFlood≥AmpLimitL2 ð2Þ

Distribution criteria:

avAmpDFlow>avAmphighavAmplow

or avAmpDFlood>avAmphighavAmplow

ð3Þ

Variance criteria:

AmpVarDFlow ≤VarLimit or AmpVarDFlood≤VarLimit ð4Þ

with:

AvAmpDFlow :

Arithmetic mean of amplitudes from FB2low

to FB2high

(frequencyband 2)

AvAmpDFlood:

Arithmetic mean of amplitudes from FB3low

to FB3high

(frequencyband 3)

AvAmplow :

Arithmetic mean of amplitudes from FB1low to FB1

high (frequencyband 1)

AvAmphigh:Arithmetic mean of amplitudes from FB4

lowto FB4

high(frequency

band 4)

AmpVarDFlow :

Variance of amplitudes from FB2low to FB2high (frequency band 2)

AmpVarDFlood :

Variance of amplitudes from FB3low to FB3high (frequency band 3)

In the former version (Schimmel et al. 2013b), the distributioncriteria were also used for the seismic signals, but for the detec-tion of smaller events, this method failed, since the divergencesat the frequency bands get extenuated for smaller amplitudes. Sofor a detection based on seismic signals, the amplitude criterionwithin one frequency band and the variance criteria have to met.

For the distinction of event sizes also, two different limits areused for the amplitude of the seismic signal. For small events(level 1: AmpLimitL1), the average amplitude of the seismicfrequency band (avAmpDFlow/DFlood) has to be over 0.4 μm/s,and for level 2 events, the amplitude has to be over 1.2 μm/s(AmpLimitL2).

The detection criteria for the seismic signals are shown by thefollowing formulae:

Amplitude criteria—level 1:

avAmpDFlow=DFlood > AmpLimit1 ð5Þ

Amplitude criteria—level 2:

avAmpDFlow=DFlood > AmpLimit2 ð6Þ

Variance criteria:

AmpVarDFlow=DFlood≤VarLimit1 ð7Þ

with:

AvAmpDFlow=flood :

Arithmetic mean of amplitudes from FB2low to FB2high (frequencyband 2)

AmpVarDFlow=flood :Variance of amplitudes from FB2

lowto FB2

high(frequency band 2)

All these criteria have to be met independent of theinfrasound and the geophone signals, within a specific timespan which is set to 12 s to report an event identification andsend out a detection signal. This combination of all the detec-tion criteria based on both signal types results in a strongreduction of false alarms. The sequence of this detection algo-rithm, which is calculated every second, is shown as a flowchart in Fig. 3, and Fig. 4 presents this detection principledepicted in running spectra of a debris flow.

The limits and time span for the detection algorithm havebeen determined in the analyzing processes of different debrisflows/debris floods and interfering signals, and they are cur-rently evaluated at several test sites. Frequent fine tuning ofthe algorithm and change in the parameters were applied toreduce the number of false alarms and to increase the detec-tion probability within early detection times even for smallevents. The current settings for the frequency bands, the limits,and the time span for detection are listed in Table 2. Theamplitude limits for level 1 and level 2 are chosen in thisversion aiming to detect events as soon as possible neglectingwhether the events are debris flows or debris floods and evenensure the detection of small debris floods events. In a futureapplication as warning system, these limits should be adaptedto the local conditions (background noise, required early warn-ing times, minimum event size for warning, etc.) and might beset to higher values.

The applicability of this detection algorithm is shown inSection 3 in the examples of a debris flow at the Lattenbach testsite and debris floods at the Dristenau and Farstrinne test sites.

Original Paper

Landslides 13 & (2016)1184

Fig. 3 Flow chart of the detection algorithm

Landslides 13 & (2016) 1185

Detection examples

Debris flow—test site LattenbachFor this example, the described approach for the detectionalgorithm has been applied to the seismic and infrasoundsignals of a debris flow at Lattenbach. The Lattenbach torrentin Tyrol (catchment 5.3 km2) is an observation site for debrisflows, operated by the Institute of Mountain Risk Engineering(BOKU, Vienna) and has been equipped with infrasound andseismic sensors since 2004. Three monitoring stations areinstalled along the torrent which are equipped with radargauges, geophones, and video cameras. Since July 2012, thedetection system consisting of a Chaparral infrasound sensorand a SM-4 geophone (changed in 2014 to a SG-5 geophone)is installed at the test site near the monitoring station 2. Anoverview of the test site with the location of the monitoringstation and the catchment area is shown in Figs. 5, and adetailed map of the monitoring setup at station 2 is given in

Fig. 6. The registered infrasound and seismic signals areverified by the flow stage level recorded with the radar gaugethat is also placed at the monitoring station 2 which startsrecording when the flow height reaches 30 cm.

This debris flow event, used as an example for the detectionalgorithm, was recorded on 1 September 2008 and had a peakdischarge of 380 m3/s and a total volume of 14,000 m3. Detailedinformation of the debris flow and the recorded infrasound andseismic data can be found at Kogelnig et al. (2012). Since this eventoccurred before the installation of the new detection system, thedata was collected with a Campbell Scientific CR1000 data logger,and the detection algorithm was applied to this data afterwards.Figure 7 shows the infrasound and seismic data of this debris flowand the level measured by the radar gauge. In the time series ofboth sensors, the arrival of the debris flow is characterized by asudden increase in amplitudes at 650 s (Fig. 7a, b). The two parallellines over the diagrams indicate the points in time of the firstdetection for the different detection levels.

Table 2 Current settings for the detection algorithm

Infrasound signal Seismic signal

Frequency band 1 FB1low–FB1high 3 to 5 Hz –

Frequency band 2—debris flow FB2low–FB2high 5 to 15 Hz 10 to 30 Hz

Frequency band 3—debris flood FB3low–FB3high 15 to 35 Hz 10 to 30 Hz

Frequency band 4 FB4low–FB4high 35 to 50 Hz –

Limit for amplitudes—level 1 AmpLimitL1 10 mPa 0.4 μm/s

Limit for amplitudes—level 2 AmpLimitL2 30 mPa 1.2 μm/s

Limit for variance VarLimit 0.6 0.6

Time span for detection Tdet 12 s 12 s

Fig. 4 Detection principle depicted in a running spectrum of a debris flow

Original Paper

Landslides 13 & (2016)1186

The maximum amplitudes of the infrasound signals producedby the debris flow are up to 5 Pa, and the maximum seismicamplitudes are up to 2×10−3 m/s. As demonstrated in Kogelniget al. (2012), wave packages corresponding to four surges of thedebris flow can be identified in the time series between 650 and800 s (Fig. 7a, b). Both signals present a spindle shape in the timeseries. The total duration of the debris flow signal in the seismicand the infrasound data is 1,650 s (650 to 2,300 s). The averageamplitudes of the four frequency bands (Fig. 7c) show that theinfrasound signal of the debris flow has its peak frequencies in the5- to 15-Hz band. The running spectra of the debris flow (Fig. 7e, f)

show a similar signal pattern in the seismic and infrasonic data.Both have a spindle shape with a rather sudden increase in fre-quencies, and energy as the debris flow approaches the sensorlocation. The frequency content slowly decreases again in bothsensors when the debris flow moves downstream far from themonitoring station. The event will be registered based on theseismic signal at second 207, while it will be identified by theinfrasound data at 539 s for level 1 and at 636 s for level 2. So atthis setup, the detection of the debris flow will be determined bythe infrasound signal, and the combination of both sensors willdetect the debris flow first 111 s before it passes the sensor site (at650 s) for the level 1 alarm, which is an adequate time for earlywarning and 14 s before passing for the level 2 alarm.

Debris flood—test site DristenauThe Dristenau torrent is located in the northern limestone alpsnear Pertisau in Tyrol. The catchment area is about 9.9 km2 withfrequent sediment transport and debris floods. Two monitoringstations (Fig. 8) equipped with precipitation gauge, video camera,and radar are installed in the upper part of the torrent. Thedetection system provided with a MK-224 infrasound sensor anda SM-4 geophone (2013) and a SG-5 geophone (2014), respectively,have been installed in June 2013 closed to monitoring station 1 asshown on the map in Fig. 9.

The debris flood used in this example occurred on 9 July 2013 at3 pm with a maximum flow height of 57 cm. The infrasound andthe seismic signals and the flow depth are presented in Fig. 10. Themaximum amplitudes of the infrasound signals are up to 400 mPa,and the maximum seismic amplitudes are up to 400×10−7 m/s. Theduration of the debris flood identified in the time series of theseismic and infrasound signals is approximately 5,000 s (2,000 to7,000 s).

In the time series of the infrasound signal, several highamplitude peaks are observed in the interval (0 to 2,000 s)(Fig. 10a). Similar peaks but with smaller amplitude are

Fig. 6 Map of the monitoring setup at the Lattenbach torrent (background imagesource: Bing maps)

Fig. 5 Overview catchment area and monitoring stations at the Lattenbach torrent (a) (background image source: Google maps) and the place of the detection system (b)

Landslides 13 & (2016) 1187

Fig. 7 Infrasound and seismic data of a debris flow monitored at the Lattenbach test site on 1 September 2008. Signals are represented with a common base of time. aInfrasound time series. b Seismogram. c Average amplitude of the four frequency bands of the infrasound signal. d Average amplitude of the frequency band of theseismic signal. e Running spectrum of the infrasound signal. f Running spectrum of the seismic signal. g Flow depth. Lines: time of first detection based on infrasoundand seismic data for level 1 and level 2

Original Paper

Landslides 13 & (2016)1188

observed in the seismic data as well in the same interval(Fig. 10b). As explained in Kogelnig et al. (2012), these ampli-tudes might correspond to the passing of a thunderstorm overthe area. In the time series of both sensors, a sharp increase inamplitudes at 2,000 s (Fig. 10a, b) is observed, corresponding tothe passing of the main surge of a debris flood which can alsobeen seen at the signal from the level sensor. Unfortunately, theradar gauge was not time synchronized with the detection sys-tem, so this may explain the increment of the flow height beforethe seismic and infrasound signals rise. After the passing of themain surge at 2,200 s, the infrasound signal forms again aspindle shape in the time series. The frequencies distributionin the seismic spectrum (Fig. 10f), have no characteristic distri-butions like in the debris flow signal at Lattenbach, due to therather small size of the debris flood. Looking only at the signals

in the time series, no significant difference (except the magni-tude) to the debris flow event discussed in Section 3.1 can beidentified. However, the frequency distributions of theinfrasound signal in the running spectrum (Fig. 10e, f) or theaverage amplitudes of the four frequency bands (Fig. 10c) revealthe difference.

The infrasound signals have peak frequencies in the 15- to35-Hz band, whereas for the debris flow event discussed pre-viously, the peak frequencies range from 5 to 15 Hz. Thesevalues hint that the characteristics of the processes must bedifferent. This difference in the frequency distribution is clearin the figure of the total spectrum of the debris flow atLattenbach and the debris flood at Dristenau (Fig. 11). So thefrequency of the main amplitude of the infrasound signal cangive information about the difference in the processes, and itcould be used to distinguish between debris flows and debrisfloods. The detection algorithm recognized the first level 1detection based on both infrasound and seismic signals on389 s after the start of recording, which was the first smallsurge with a duration of about 20 s. The main event wasdetected based on both signals at 2,004 s for level 1 and laterat 2,083 s for level 2. That means the event could only bedetected after the surge passes the sensor site (~2,000 s). Onereason for the late detection might be the small event size,another reason is that the monitoring station is located veryclose to the valley end where debris flows or debris floods aregenerated, so that there is only a short distance for a detectionin advance. This shows that the event magnitude and thegeographical conditions have high impact to the detectionand early warning times.

Debris flood—test site FarstrinneThe torrent Farstrinne is a new test site of the Institute ofMountain Risk Engineering and is located near Umhausen in theÖtztal in Tyrol. The catchment area of this site is about 5.5 km2,and the detection system equipped with a Chaparral infrasoundsensor and a SG-5 geophone is now installed there since July 2013.In August 2014, a debris flow radar (Koschuch et al. 2015) and a

Fig. 9 Map of the monitoring setup at the Dristenau torrent (background imagesource: Google maps)

Fig. 8 Overview catchment area and monitoring stations at the Dristenau torrent (a) (Source: Google maps) and the place of the detection system (b)

Landslides 13 & (2016) 1189

Fig. 10 Infrasound and seismic data of a debris flood monitored at the Dristenau test site on 9 July 2013. Signals are represented with a common base of time. aInfrasound time series. b Seismogram. c Average amplitude of the four frequency bands of the infrasound signal. d Average amplitude of the frequency band of theseismic signal. e Running spectrum of the infrasound signal. f Running spectrum of the seismic signal. g Flow depth. Lines: time of first detection based on infrasoundand seismic data for level 1 and level 2

Original Paper

Landslides 13 & (2016)1190

video camera have been also installed there. Figure 12 gives anoverview of the test site with the location of the monitoring stationand the catchment area, and Fig. 13 presents a detailed map of thesetup.

The event took place in the night of 30 July 2014 to 31 July 2014,whereby the first debris flood occurred at 23:08, and afterwards, asecond event occurred few hours later at 5:00. The total dischargeof these two events was approximately 5,000 m3, and they stoppedat the upper end of the basin. Unfortunately, the debris flow radarwas not operated at the time of the events, so there was nopossibility to verify the exact time when this event occurred. Butpictures before and after the night of the events showed a shiftingof the torrent from the left (Fig. 14a) to the right side (Fig. 14b) andthe deposit at the retention basin verified the event surelyoccurred.

From the frequency range of the infrasound signal, these eventsseem more likely to be debris floods, since the main amplitudesare located in 15 to 35 Hz band, which indicates a higher watercontent, but the difference from the debris flow frequency band isnot so significant like in the Dristenau event described inSection 3.2. The maximum amplitudes of the infrasound signalsproduced by the both debris floods are up to 400 mPa for the firstevent and 800 mPa for the second and the maximum seismicamplitudes are up to 300 and 500×10–7 m/s, respectively. Theinfrasound and the seismic signals of the first event are presentedin Fig. 15 and of the second event in Fig. 16.

The detection of both events was accomplished by theseismic signals only, and for the first event, the level 1 detec-tion took place at 1,759 s and the level 2 detection at 1,787 s.For the second event, they took place at 1,301 s (level 1) and

Fig. 12 Overview catchment area and monitoring stations at the Farstrinne torrent (a) (source: Google maps) and the place of the detection system (b)

Fig. 11 Total spectrum of the infrasound signal of the debris flow at Lattenbach (a) and of the debris flood at Dristenau (b)

Landslides 13 & (2016) 1191

1,353 s (level 2), respectively. So the events were detected atlevel 1 at 21 and 99 s, respectively, before the main surgepassed through the sensor site (first event 1,780 s, secondevent 1,400 s). If the detection based only on the infrasoundsignal were accomplished earlier, warning times of 98 and170 s, respectively, would be possible, but this would resultin more false alarms. The limitation for the early detectiontime caused by the seismic detection might be a result of thenot optimal selected geophone position in the unconsolidatedsoil at the top of the dam, and so, the detection time couldbe increased by a better seismic setup (e.g., fixing the geo-phone in the concrete of the check dam). The comparison ofthe results of the detection algorithm for the events atLattenbach, Dristenau, and Farstrinne shows significant differ-ences between the detection times based on seismic orinfrasound data depending on the signal sequence, magnitude,and local conditions that could lead to a wide variance of theearly warning times.

DiscussionThis paper presents a first approach for a debris flow/debris flooddetection system based on a combination of seismic andinfrasound sensors. The combination of both sensor technologiesincreases the detection probability and minimizes false alarms. Forexample at the test site Farstrinne, 176 detections based only onseismic data and six detections based only on infrasound data(analysis per hour) have been registered in the 2014 season; how-ever, the combination of both sensors leads to only two correctevent detections and no false alarm. An overview of the number ofevents at the different test sites and the detections or false alarmsfor the season 2013 is given in Table 3 and for 2014 in Table 4. Inthese tables, for the sake to compare the detection efficiency, theevents are classified depending on their maximum amplitude ofthe infrasound signal into very small, small, and medium scaleevents. At 2013 and 2014, no larger event occurred at the equippedtest sites.

All medium-sized events with infrasound amplitudes greaterthan 400 mPa have been detected successfully in both years, andeven the smaller events have been detected with high accuracy. Thefalse alarms registered at the test site Lattenbach in the year 2013have been caused by a not correctly working geophone. After thechange of the geophone at the test site Lattenbach in 2014, no falsealarm was registered any more. The false alarms at Dristenau areof unknown reason. At the detection level 2, no false alarm oc-curred in both years at the test sites Lattenbach and Dristenau. Asalready discussed, a large variation of detection times at thedifferent test sites has been identified. While for larger eventsand optimal location of the detection system the time betweendetection and passing of the main surge can be up to 1 min andmore (Lattenbach), this time can go down to zero or less forsmaller events and disadvantageous sensor sites (Dristenau). Sothe setup of the sensors has to be chosen carefully: close to thetorrent, protected from wind and other sources of interference andon consolidated soil or rock for the geophone.

ConclusionThe main objective of this work was to design a detection systemwhich is based on a minimum of one seismic and one infrasoundsensor that are co-located and a microcontroller which runs a detec-tion algorithm to detect debris flows and debris floods with highaccuracy in real time directly on-site. This system can easily be

Fig. 14 Change of the Farstrinne torrent caused by the events at the night of 30 to 31 August 2014

Fig. 13 Map of the monitoring setup at the Farstrinne torrent (background imagesource: Bing maps)

Original Paper

Landslides 13 & (2016)1192

Fig. 15 Infrasound and seismic data of the first debris flood monitored at the Farstrinne test site on 30 July 2014. Signals are represented with a common base of time. aInfrasound time series. b Seismogram. c Average amplitude of the four frequency bands of the infrasound signal. d Average amplitude of the frequency band of theseismic signal. e Running spectrum of the infrasound signal. f Running spectrum of the seismic signal. Lines: time of first detection based on infrasound and seismic datafor level 1 and level 2

Landslides 13 & (2016) 1193

Fig. 16 Infrasound and seismic data of the second debris flood monitored at the Farstrinne test site on 31 July 2014. Signals are represented with a common base of time.a Infrasound time series. b Seismogram. c Average amplitude of the four frequency bands of the infrasound signal. d Average amplitude of the frequency band of theseismic signal. e Running spectrum of the infrasound signal. f Running spectrum of the seismic signal. Lines: time of first detection based on infrasound and seismic datafor level 1 and level 2

Original Paper

Landslides 13 & (2016)1194

enhanced to be a quick and simple to install low-cost warning system.First tests of the detection system in 2013/2014 at five test sites inAustria showed already promising results. Although no larger debrisflows occurred in that period within these test sites, the system hasdetected all medium-sized events (nine events) with infrasound am-plitudes greater than 400 mPa and also over 70 % of smaller events,while only seven false alarms were registered. However, the applica-tion of seismic and infrasound sensors for monitoring and detectionof alpine mass movements is not a straightforward task.Understanding the propagation and the attenuation mechanism ofseismic and infrasonic waves and the background noise characteristicsunder the studied conditions are crucial for the interpretation of therecorded seismic and infrasonic signals and for the development ofthe detection algorithm. The equipment and the placement of thesensors have to be chosen carefully. Also, the settings of the detectionalgorithm (amplitude limits) have to be adjusted appropriate to theapplication of the system and the local conditions of the site. So, if thesite is exposed to wind, the detection probability will be reduced and asite with too much artificial noise (cars, trains, power plants, etc.)might need higher amplitude limits for the detection algorithm, whichwill also reduce the detection probability and early warning times, ortoo much false alarms can be registered. In summary, the analysesconfirmed that debris flows and debris floods produce seismic andinfrasonic signals with the characteristics that are reproducible at verydifferent experimental sites and environmental conditions. Therefore,for the monitoring of alpine mass movements, the combination of

infrasound and seismic sensors and using these signals for an auto-matic detection should lead to the promising result.

References

Arattano M. Monitoring the presence of the debris-flow front and its velocity throughground vibrations detectors. Proceedings of the Third International Conference onDebris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment, 731–743,Millpress, Rotterdam (2003)

Bedard A.J. Detection of avalanches using atmospheric infrasound. Proceedings: WesternSnow Conference, Fort Collins, CO

Bedard A.J. (1994) An evaluation of atmospheric infrasound for monitoring avalanches.Proceedings: 7th international symposium on acoustic sensing and associated tech-niques of the atmosphere and oceans, Boulder, CO

Bedard A.J. (1996) Infrasonic and near infrasonic atmospheric sounding and imaging,NOAA/EARL/Environmental Technology Laboratory

Biescas B, Dufour F, Furdada G, Khazaradze G, Suriñach E (2003) Frequency contentevolution of snow avalanche seismic signals. Surv Geophys 24(5–6):447–464

Chou H.T., Cheung Y.L., Zhang S.C. (2007) Calibration of infrasound monitoring systemsand acoustic characteristics of debris-flow movements by field studies. Institute ofMountain Hazards and Environment, Chinese Academy of Science and Ministry ofWater resources

Chou, H.T., Chang, Y.L. and Zhang, S.X. (2010) Acoustic signals and geophone responseof rainfall-induced debris flows. J Chin Inst Eng

Coviello V, Arattano M, Turconi L (2015) Detecting torrential processes from a distancewith a seismic monitoring network. Nat Hazards 78(3):2055–2080

Huang C., Yin H., Shieh C. (2003) Experimental study of the underground soundgenerated by debris flow. Proc. of the Third Int. Conference on Debris-Flow Hazards

Table 3 Overview of events and detections in 2013

Test site Number of events Detections False alarmsSize of event: Very small

<100 mPaSmall

>100 mPa<400 mPa

Medium>400 mPa

Very small<100 mPa

Small>100 mPa<400 mPa

Medium>400 mPa

Lattenbach 0 0 0 0 0 0 4

Wartschenbach 8 3 1 2 2 1 0

Farstrinne 0 0 0 0 0 0 0

Dristenau 18 4 4 12 3 4 2

Schüsserbach 2 3 1 0 2 1 0

50 % 70 % 100 % 6

Table 4 Overview of events and detections in 2014

Test site Number of events Detections False alarmsSize of event: Very small

<100 mPaSmall

>100 mPa<400 mPa

Medium>400 mPa

Very small<100 mPa

Small>100 mPa<400 mPa

Medium>400 mPa

Lattenbach 3 0 0 2 0 0 0

Farstrinne 0 0 2 0 0 2 0

Dristenau 10 1 1 4 1 1 1

Schüsserbach 0 0 0 0 0 0 0

45 % 100 % 100 % 1

Landslides 13 & (2016) 1195

Mitigation: Mechanics, Prediction and Assessment, Vol.2, 743–753. Millpress,Rotterdam

Huang C-J, Yin H-Y, Chen C-Y, Yeh C-H, Wang C-L (2007) Ground vibrations produced byrock motions and debris flows. J Geophys Res: Earth Surf 112:F02014. doi:10.1029/2005JF000437

Hübl J, Schimmel A, Kogelnig A, Suriñach E, Vilajosana I, McArdell BW (2013) A review onacoustic monitoring of debris flow. Int J Saf Sec Eng 3(2):105–115, ISSN 2041–9031

Kogelnig A., Hübl J., Suriñach E., Vilajosana I., McArdell B.W.(2010) Infrasound produced bydebris flow: propagation and frequency content evolution. Online first, Nat. Hazards

Kogelnig A.(2012) Development of acoustic monitoring for alpine mass movements. PhDThesis, University of Natural Resources and Life Sciences (BOKU), Vienna, Institute ofMountain Risk Engineering

Kogelnig A.(2008) Infrasound monitoring of gravity driven mass movements: avalanches anddebris flow. In: European Geosciences Union (Ed.), Geophysical Research Abstracts, Vol.10, EGU2008-A-08537, EGU General Assembly 2008, Wien, ISSN 1029–7006

Kogelnig A., Hübl J.(2009) Infrasound monitoring of debris flow at Lattenbach, Austria.In: European Geosciences Union (Ed.), Geophysical Research Abstracts, Vol. 11,EGU2009-2573, EGU General Assembly 2009, Wien

Koschuch R, Jocham P, Hübl J (2015) One year use of high-frequency radar technology inalpine mass movement monitoring: principles and performance for torrential activ-ities. Eng Geol Soc Territor 3:69–72

Marchi L, Arattano M, Deganutti A (2002) Ten years of debris-flow monitoring in theMoscardo Torrent (Italian Alps). Geomorphology 46(1–2):1–17

Pilger C, Bittner M (2009) Infrasound from tropospheric sources: impact on mesopausetemperature? J Atmosph Solar-Terrestr Phys 71:816–822

Rabiner LR, Schafer RW, Rader CM (1969) The chirp z-transform algorithm and itsapplication. Bell Syst Tech J 48:1249–1292. doi:10.1002/j.1538-7305.1969.tb04268.x

Schimmel A; Hübl J.(2013) Automatic detection of avalanches using infrasound andseismic signals. Naaim-Bouvet, F; Durand, Y; Lambert, R (Eds.), Proceedings ofISSW2013, 904–908

Schimmel A; Hübl J.(2013) Development of a debris flow warning system based on acombination of infrasound and seismic signals. Rickenmann D; Laronne J.B.; TurowskiJ.M.; Vericat D (Eds.), Abstracts, 98–99

Scott ED (2004) Results of recent infrasound avalanche monitoring studies. Proceedings:International Snow Science Workshop, Jackson Hole, Wyoming

Scott ED (2006) Practical implementation of avalanche infrasound monitoring technologyfor operational utilization near Teton Pass Wyoming. Proceedings: International SnowScience Workshop

Scott E, Hayward C, Kubichek R, Hamann J, Comey R, Pierre J, Mendenhall T (2007)Single and multiple sensor identification of avalanche generated infrasound. Cold RegSci Technol 47:159–170

Sommerfeld RA (1977) Preliminary observations of acoustic emissions preceding ava-lanches. J Glaciol 19(81):399–409

Sommerfeld RA, Gubler H (1983) Snow avalanches and acoustic emissions. Ann Glaciol 4,Int Glaciol Soc

Suriñach E, Sabot F, Furdada G, Vilaplana J (2000) Study of seismic signals of artificiallyreleased snow avalanches for monitoring purpose. Phs Chem Earth B 25(9):721–727

Suriñach E., Kogelnig A., Vilajosana I., Hübl J., Hiller M., Dufour F (2009) Incoporación della señal de infrasonido a la detección y estudio de aludes de nieve y flujostorrenciales,VII SimposioNacinal sobre Taludes y LaderasInestables, Barcelona, Spain

Vilajosana I, Suriñach E, Abellán A, Khazaradze G, Garcia D, Llosa J (2008) Rockfallinduced seismic signals: case study in Montserrat, Catalonia. Nat Hazards Earth SystSci 8(4):805–812

Wu J, Kang Z, Tian L, Zhang S (1990) Observation and investigation of debris flows atJiangjia Gully in Yunnan Province (China). Sci Press, Beijing

Zhang S., Hong Y., Yu B (2004) Detecting infrasound emission of debris flow for warningpurpose, 10. Congress Interpraevement

A. Schimmel ()) : J. HüblInstitute of Mountain Risk Engineering,University of Natural Resources and Life Sciences,Vienna, Austriae-mail: andreas.schimmel@boku.ac.at

Original Paper

Landslides 13 & (2016)1196